Swedish vaccine trials have been used to examine sensitivity and specificity of diagnostic procedures for Bordetella pertussis infection. The proportions of cases diagnosed by culture and serology were 55% and 45%, respectively, when both methods were optimized. The culture method included nasopharyngeal aspiration, direct inoculation on plates, enrichment, and repeated collection of samples. An enzyme-linked immunosorbent assay for IgG antibodies to pertussis toxin (PT) and to filamentous hemagglutinin, with paired sera, was used for serology. Preexposure sera other than the acute serum increased the sensitivity of serology by 10%. A serology quality-assurance program to control imprecision and allow comparability over time and between laboratories is described. The direct fluorescent antibody technique had a sensitivity of 38% and a specificity of 99.6% in comparison with culture. A nested polymerase chain reaction (PCR) with the PT promoter region as target was 95% sensitive in comparison with culture if a cation-exchange resin was used to reduce inhibition. PCR enabled us to identify 83 positive samples in addition to 215 culture-positive ones-an increase of 38%--all with other indicators of pertussis infection.